Fig 1.
Summary of available occurrence records across all species.
Maps displaying available occurrence records based on: (a) a 10 km raster grid considering all records; (b) a 1km raster using only records with resolution of 1km or higher (aggregated to 10km for comparison); combined across all species to give: (i) total recording effort; (ii) species richness. (c) maps the loss of information incurred by considering a higher precision model. This highlights the emergence of regional data “deserts” where the resolution of records has been limited by local providers.
Fig 2.
Summary of available density estimates across all species.
Maps displaying extracted density estimates combined across all species to give: (a) total surveying effort (the number of density estimates in each square, with some species contributing more than a single estimate); (b) species richness (the number of species for which a usable density estimate exists); and (c) recording recentness (decade of last survey). Values derived from original polygon representations by converting to 25m raster grids and then aggregating cells to a display resolution of 10km.
Fig 3.
Diagram of the model framework.
Outline of modelling process for each species (maps show output for the hedgehog: Erinaceus europaeus; for detailed discussion of these results refer to the species specific report in S6 File). Initially, occurrence is coupled with environmental data comparing 7 SDMs to identify the “best” habitat suitability map based on AUC. This is repeated 100 times with the best maps combined to produce an overall mean habitat suitability; the mid value indicates the threshold above which occurrence is assumed. For these cells, habitat suitability scores are then matched with extracted density estimates and linear regression performed to predict abundance.
Fig 4.
Plot (a) presents the available data showing: (i) species occurrence; and (ii) density; by the decade of last sighting; (iii) mean density (estimates are assumed to be representative of entire cell, considered the upper limit of observed density). Plot (b) presents the corresponding model predictions based on a 10 km grid showing: (i) habitat suitability; (ii) the lower bound (Minimum); and (iii) the upper bound (Maximum); for abundance. Maps show the fox is a highly reported and surveyed species. Habitat suitability modelling reflects this predicting widespread occurrence across a variety of landscapes. Maps of abundance show similar spatial distributions suggesting a negative correlation between habitat suitability and density. Uncoloured areas indicate absence (i.e. zero abundance) which is assumed where habitat suitability scores are lower than a threshold value; in this case 0.9.
Table 1.
Summary of observed data and model predictions by land cover for the red fox.
Table 2.
Summary of model predictions and analysis of observed data for the 36 “most” surveyed species.
Fig 5.
Summary of model predictions across all species.
Maps displaying model outputs based on: (a) a 10 km raster grid considering all records; (b) a 1km raster using only records with resolution of 1km or higher (aggregated to 10km for comparison); combined across all species to give: (i) predicted species richness; (ii) Minimum predicted biomass; (iii) Maximum predicted biomass. There is good agreement between predicted richness and that observed in Fig 1. However, there is a degree of inconsistency, particularly based on a 10km raster, between the spatial distributions of minimum and maximum biomass.